1. Field of the Invention
The present invention relates in general to memory array repair and, in particular, to devices and methods for repairing memory arrays, such as dynamic random access memory arrays, by storing each individual bit in multiple memory cells in the arrays.
2. State of the Art
In general, Dynamic Random Access Memory (DRAM) arrays store digital information in the form of “1” and “0” bits by storing the bits as electric charges on capacitors. DRAM arrays then retrieve the stored bits by discharging their representative electric charges to a conductor, such as a digit line, and then detecting a change in voltage on the conductor resulting from the discharge. When any of the capacitors in a DRAM array are unable to store a sufficient electric charge to cause a detectable change in voltage on a conductor when discharged to the conductor, any “1” or “0” bits stored as electric charges on those capacitors cannot be retrieved by the array. In this circumstance, the array must be repaired by replacing the failing capacitors with redundant capacitors in redundant rows or columns in the array. If too many of the redundant capacitors also fail, then the array must be discarded.
More specifically, a conventional DRAM array 10 shown in FIG. 1 stores digital information in the form of “1” and “0” bits by storing the bits as electric charges on storage capacitors 12, 14, and 16 in memory cells 18, 20, 22 and 24 arranged along word lines (i.e., rows) WL0, WL1, . . . , and WLm and complementary pairs of digit lines (i.e., columns) D0 and D0*, D1 and D1*, . . . , and Dn and Dn*. Of course, while the DRAM array 10 is shown in FIG. 1 as having only nine memory cells in order to simplify description, the DRAM array 10 typically includes thousands or millions of memory cells.
The DRAM array 10 stores a “1” bit in the memory cell 18, for example, by energizing the word line WL0 to activate an NMOS transistor 26. The DRAM array 10 then applies a “1” bit voltage equal to a supply voltage VCC (e.g., 3.3 Volts) to the digit line D0, causing current to flow from the digit line D0, through the activated NMOS transistor 26 and the storage capacitor 12, and to a cell plate voltage DVC2 typically equal to one half the supply voltage VCC. As this current flows, the storage capacitor 12 stores positive electric charge received from the digit line D0, causing a voltage VS1 on the storage capacitor 12 to increase. When the voltage VS1 on the storage capacitor 12 equals the “1” bit voltage on the digit line D0, current stops flowing through the storage capacitor 12. A short time later, the DRAM array 10 de-energizes the word line WL0 to de-activate the NMOS transistor 26 and isolate the storage capacitor 12 from the digit line D0, thereby preventing the positive electric charge stored on the storage capacitor 12 from discharging back to the digit line D0.
Similarly, the DRAM array 10 stores a “0” bit in the memory cell 20, for example, by energizing the word line WL1 to activate an NMOS transistor 28. The DRAM array 10 then applies a “0” bit voltage approximately equal to a reference voltage VSS (e.g., 0.0 Volts) to the digit line D0, causing current to flow from the cell plate voltage DVC2, through the storage capacitor 14 and the activated NMOS transistor 28, and to the digit line D0. As this current flows, the storage capacitor 14 stores negative electric charge received from the digit line D0, causing a voltage VS2 on the storage capacitor 14 to decrease. When the voltage VS2 equals the “0” bit voltage on the digit line D0, current stops flowing through the storage capacitor 14. A short time later, the DRAM array 10 de-energizes the word line WL1 to de-activate the NMOS transistor 28 and isolate the storage capacitor 14 from the digit line D0, thereby preventing the negative electric charge stored on the storage capacitor 14 from discharging back to the digit line D0.
The DRAM array 10 stores “1” and “0” bits in the memory cells 22 arranged along the complementary digit lines D0*, D1*, . . . , and Dn* in a manner similar to that described above, with the exception that the “1” bit voltage for these cells is approximately equal to the reference voltage VSS and the “0” bit voltage equals the supply voltage VCC.
The DRAM array 10 retrieves “1” and “0” bits stored in the manner described above in the memory cells 18, 20, 22, and 24 by discharging electric charges stored on the storage capacitors 12, 14, and 16 to the digit lines D0, D0*, D1, D1*, . . . , Dn, and Dn* and then detecting a change in voltage on the digit lines D0, D0*, D1, D1*, . . . , Dn, and Dn* resulting from the discharge with sense amplifiers (0), (1), . . . , and (n).
For example, the DRAM array 10 retrieves the “1” bit stored in the memory cell 18 by first equilibrating the voltages on the digit lines D0 and D0* to the cell plate voltage DVC2. The DRAM array 10 then energizes the word line WL0 to activate the NMOS transistor 26, causing the positive electric charge stored on the storage capacitor 12 to discharge through the activated NMOS transistor 26 to the digit line D0. As the positive electric charge discharges, the voltage on the digit line D0 rises by an amount VSENSE calculated as follows:VSENSE=(VS−DVC2)*CS/(CD+CS)where VS is the voltage VS1 on the storage capacitor 12, CS is the capacitance of the storage capacitor 12, and CD is the capacitance of the digit line D0. When the rise in voltage VSENSE on the digit line D0 causes a difference in voltages between the digit lines D0 and D0* to exceed a detection threshold (typically about 150 mVolts) of the sense amplifier (0), the sense amplifier (0) responds by driving the voltage on the digit line D0 to the supply voltage VCC and by driving the voltage on the digit line D0* approximately to the reference voltage VSS. Input/output gating circuitry, DC sense amplifiers, and an output buffer (not shown) then transmit these voltages from the digit lines D0 and D0* to external circuitry as a “1” bit.
Likewise, the DRAM array 10 retrieves the “0” bit stored in the memory cell 20, for example, by first equilibrating the voltages on the digit lines D0 and D0* to the cell plate voltage DVC2. The DRAM array 10 then energizes the word line WL1 to activate the NMOS transistor 28, causing the negative electric charge stored on the storage capacitor 14 to discharge through the activated NMOS transistor 28 to the digit line D0. As the negative electric charge discharges, the voltage on the digit line D0 falls by an amount VSENSE, calculated as described above, where VS is the voltage VS2 on the storage capacitor 14 and CS is the capacitance of the storage capacitor 14. When the drop in voltage VSENSE on the digit line D0 causes the difference in voltages between the digit lines D0 and D0* to exceed the detection threshold of the sense amplifier (0), the sense amplifier (0) responds by driving the voltage on the digit line D0 approximately to the reference voltage VSS and by driving the voltage on the digit line D0* to the supply voltage VCC. The input/output gating circuitry, DC sense amplifiers, and output buffer then transmit these voltages from the digit lines D0 and D0* to external circuitry as a “0” bit.
The DRAM array 10 retrieves “1” and “0” bits from the memory cells 22 arranged along the complementary digit lines D0*, D1*, . . . , and Dn* in the same manner as described above.
DRAM arrays sometimes contain defective memory cells that are unable to reliably store “1” and “0” bits in the manner described above. In some instances, this occurs because the capacitance of the storage capacitors in these memory cells is too small, preventing the capacitors from retaining a sufficient electric charge to cause a change in voltage VSENSE on a digit line when discharged to the digit line that exceeds a sense amplifier's detection threshold. In other instances, this occurs because the electric charge stored on the storage capacitors in these memory cells leaks away through a variety of mechanisms, also preventing the capacitors from retaining a sufficient electric charge to cause a detectable change in voltage VSENSE on a digit line when discharged to the digit line. In either case, because the change in voltage VSENSE caused by discharging the electric charges stored by the storage capacitors in these memory cells cannot be detected by a sense amplifier, the “1” and “0” bits represented by the electric charges stored in these memory cells are irretrievable.
DRAM arrays are also sometimes unable to reliably store “1” and “0” bits in certain memory cells because the detection threshold of the cells' associated sense amplifier is too large for the sense amplifier to detect a change in voltage VSENSE caused by one of the cells discharging to the digit line. In this case as well, the “1” and “0” bits represented by the electric charges stored in these cells are irretrievable.
Generally, when DRAM arrays are found to be defective in this way, an attempt is made to repair the arrays by replacing defective memory cells and defective sense amplifiers in the arrays with redundant memory cells provided in redundant rows or columns in the arrays and with redundant sense amplifiers provided in the redundant columns.
Conventionally, when a redundant row is used to repair a DRAM array containing a defective memory cell, a row address that identifies the defective cell's row is permanently stored in the array by blowing selected fuses or anti-fuses in the array. Then, during normal operation of the DRAM array, if the array receives a request to access a memory cell having a memory address including a row address portion that corresponds to the stored row address, redundant circuitry in the array directs the array to access a redundant memory cell in the redundant row instead of accessing the memory cell identified by the received memory address. Since every memory cell in the defective cell's row has the same row address, every cell in the defective cell's row, both operative and defective, is replaced by a redundant memory cell in the redundant row.
Similarly, when a redundant column is used to repair a DRAM array containing a defective memory cell, a column address that identifies the defective cell's column is permanently stored in the array by blowing selected fuses or anti-fuses in the array. Then, during normal operation of the DRAM array, if the array receives a request to access a memory cell having a memory address including a column address portion that corresponds to the stored column address, redundant circuitry in the array directs the array to access a redundant memory cell in the redundant column instead of accessing the memory cell identified by the received memory address. Since every memory cell in the defective cell's column has the same column address, every cell in the defective cell's column, both operative and defective, is replaced by a redundant memory cell in the redundant column.
The process described above for repairing a DRAM array using redundant rows and column is well known in the art, and is described in various forms in U.S. Pat. Nos. 4,459,685, 4,601,019, 5,422,850 and 5,528,539.
Because the conventional repair process described above uses an entire redundant row or column to repair each defective memory cell in a DRAM or other memory array, the number of defective memory cells that can be repaired in an array is limited by the number of redundant rows or columns in the array which, in turn, is limited by the space available in the array for redundant rows or columns. As a result, it is not uncommon for defective DRAM and other memory arrays to be discarded because the conventional repair process cannot repair the quantity of defective memory cells they contain. This problem is often exacerbated by the discovery of defective redundant memory cells in the available redundant rows or columns. Obviously, it would be preferable to be able to repair these defective arrays rather that discard them.
Therefore, there is a need in the art for an improved device and method for repairing DRAM arrays and other memory arrays. Such an improved device and method should be applicable to arrays containing memory cells that are unable to reliably store “1” and “0” bits because the storage capacitors do not retain a sufficient electric charge, or because their associated sense amplifiers have a detection threshold that is too large.